Bootstrapped training of event extraction classifiers
Conference Paper
Overview
Identity
Additional Document Info
View All
Overview
abstract
2012 Association for Computational Linguistics. Most event extraction systems are trained with supervised learning and rely on a collection of annotated documents. Due to the domain-specificity of this task, event extraction systems must be retrained with new annotated data for each domain. In this paper, we propose a bootstrapping solution for event role filler extraction that requiresminimal human supervision. We aim to rapidly train a state-of-The-Art event extraction system using a small set of "seed nouns" for each event role, a collection of relevant (in-domain) and irrelevant (outof- domain) texts, and a semantic dictionary. The experimental results show that the bootstrapped system outperforms previous weakly supervised event extraction systems on the MUC-4 data set, and achieves performance levels comparable to supervised training with 700 manually annotated documents.
